- [[linear models in transformed space are non-linear models in original space]], [[linear classifiers]]
# Idea
Kernel SVMs are computationally efficient algorithms that fit non-linear boundaries using [[linear classifiers]]. They exploit the fact that fitting [[linear classifiers]] in a transformed space corresponds to fitting non-linear models in the original space (see [[linear models in transformed space are non-linear models in original space|here]]). Kernels help to perform feature transformations in a computationally efficient way.
# Python code
```python
from sklearn.svm import SVC
svm = SVC(gamma=1) # default is kernel='rbf'; larger gamma, more complex boundaries
```
![[Pasted image 248.png]]
# References
- https://campus.datacamp.com/courses/linear-classifiers-in-python/support-vector-machines?ex=4